Mixed Strategy — Definition, Formula & Example
A mixed strategy is a plan in which a player randomizes, choosing each available action with a set probability instead of always picking the same one. Bluffing in poker and varying penalty-kick directions are mixed strategies.
For a player with a finite set of pure strategies , a mixed strategy is a probability distribution on : a vector with and . A pure strategy is the special case where one . By Nash's theorem (1950), every finite game has at least one Nash equilibrium in mixed strategies.
Key Formula
Where:
- = Row player's expected payoff when both players use mixed strategies
- = Probability the row player chooses row
- = Probability the column player chooses column
- = Row player's payoff when row meets column
Worked Example
Problem: In a zero-sum game, Row picks Up or Down and Column picks Left or Right. Row's winnings, which Column pays, are: Up/Left $3, Up/Right −$1, Down/Left −$2, Down/Right $4. No cell is a saddle point, because Row's maximin of −$1 differs from Column's minimax of $3. Find the equilibrium mixed strategies and the value of the game.
Step 1: Let Row play Up with probability . Row picks so that Column gets the same expected result from Left and from Right. Then Column cannot exploit Row.
Step 2: Let Column play Left with probability . Column picks so that Row's expected payoff is the same for Up and Down.
Step 3: Substitute into Row's expected payoff against Left. Any column gives the same result.
Answer: Row plays Up with probability and Down with probability . Column plays Left and Right with probability each. The game's value is $1 per round in Row's favor.
Why It Matters
Mixed strategies are how you solve games like rock-paper-scissors or matching pennies, which have no pure-strategy equilibrium. In operations research and game theory courses, two-player zero-sum games are often solved by finding optimal mixed strategies with linear programming. The same idea shapes randomized audit schedules and security patrol routes.
Common Mistakes
Mistake: Choosing your own probabilities so that your own two options pay the same.
Correction: In equilibrium, your probabilities make your opponent indifferent. That is why Row's comes from Column's payoffs in Step 1, and Column's comes from Row's payoffs in Step 2.
